Date: Sun, 27 May 2001 20:01:18 -0700 From: Ed Hudson <elh_fbsd@spnet.com> To: hackers@FreeBSD.ORG Subject: Re: general speed differences between 4.1.1-RELEASE and 4.3-RELEASE Message-ID: <200105280301.UAA48350@spnet.com> In-Reply-To: Message from Rik van Riel <riel@conectiva.com.br> of "Sun, 27 May 2001 22:25:45 -0300." <Pine.LNX.4.21.0105272222250.1907-100000@imladris.rielhome.conectiva>
next in thread | previous in thread | raw e-mail | index | archive | help
> I doubt FreeBSD would need to enable write caching in order
> to be as fast as Linux (which doesn't have write caching
i spoke too harshly.
what i meant to show is that interactive performance
is compromised under load with soft updates enabled
(although soft updates clearly speed up some general tasks and
accelerate some tasks considerably). i also i wanted
to show that hw.ata.wc=0 has 3-7x impact on fast hardware,
which is a much larger impact than almost any other single
parameter.
i had seen soft updates as a justification of turning
ata.wc off (later education on my part by the memebers of
hackers has broadened my understanding of the motivation).
i suspect that this issue was well hashed out in this news
group when i wasn't tracking the stream.
i use freebsd to help design the chips that i work on, and
i've always relied on and been impressed by its ability to
perform well handling large cad programs -
so i was just surprised at the sudden change in this default
behavior re hw.ata.wc=0. clearly, this was just ignorance
on my part, and i suspect had i looked more closely at the
release notes i would have just turned this parameter on,
kept soft-updates off and still been a happy camper.
(much kudoo's to mr. dillons now timely tuning.7, btw).
another note regarding hw.ata.wc=0 as the default -
if i assume that i've been running effectively with hw.ata.wc=1
for the last couple of years, i would extrapolate that the likelyhood
of a fbsd/ufs failure in this mode is small compared to the reliability
problems of the rest of the system, and the same protection that
covers those liabilities also cover my exposure to hw.ata.wc=1 problems
(e.g., good backups, ups's, etc).
given the huge impact that users (at least those like myself) see
of this parameter, and the reliability impact that i think i understand,
i am surprised by the choice of default. it feels like a recruiting
attempt for linux. (btw, i do think that the freebsd project is probably
the best working example of open source software, and its benefits,
so i'm not trying to promote linux - but both have benefited from
their coevolution).
(system reliability: - i think hard drive failures are maybe #1 in
occurance, motherboard and memory failures as #2, and pwr supply
failures #3, and cpu failures last.)
ok, i know the knobs to turn to solve my problems.
i'm happy. i'll shutup. thanks again to all you hackers for a great os.
i guess there really aren't evil space monsters invading the inner
sanctum...
-elh
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200105280301.UAA48350>